torch.fft: Multi-dimensional transforms #44550

peterbell10 · 2020-09-11T13:02:06Z

Part of the torch.fft work (gh-42175).
This adds n-dimensional transforms: fftn, ifftn, rfftn and irfftn.

This is aiming for correctness first, with the implementation on top of the existing _fft_with_size restrictions. I plan to follow up later with a more efficient rewrite that makes _fft_with_size work with arbitrary numbers of dimensions.

Stack from ghstack:

torch.fft: Two dimensional FFT functions #45164 torch.fft: Two dimensional FFT functions
codegen: Improve array default handing #45163 codegen: Improve array default handing
torch.fft helper functions #44877 torch.fft helper functions
Deprecate old fft functions #44876 Deprecate old fft functions
torch.fft: Multi-dimensional transforms #44550 torch.fft: Multi-dimensional transforms

Differential Revision: D23846032

[ghstack-poisoned]

dr-ci · 2020-09-11T13:18:26Z

💊 CI failures summary and remediations

As of commit 9bccee0 (more details on the Dr. CI page):

💚 💚 Looks good so far! There are no failures yet. 💚 💚

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions on the GitHub issue tracker or post in the (internal) Dr. CI Users group.

See how this bot performed.

This comment has been revised 77 times.

[ghstack-poisoned]

ghstack-source-id: bc92ef4 Pull Request resolved: #44550

Part of the `torch.fft` work (gh-42175). This adds n-dimensional transforms: `fftn`, `ifftn`, `rfftn` and `irfftn`. This is aiming for correctness first, with the implementation on top of the existing `_fft_with_size` restrictions. I plan to follow up later with a more efficient rewrite that makes `_fft_with_size` work with arbitrary numbers of dimensions. [ghstack-poisoned]

ghstack-source-id: 16ea948 Pull Request resolved: #44550

Part of the `torch.fft` work (gh-42175). This adds n-dimensional transforms: `fftn`, `ifftn`, `rfftn` and `irfftn`. This is aiming for correctness first, with the implementation on top of the existing `_fft_with_size` restrictions. I plan to follow up later with a more efficient rewrite that makes `_fft_with_size` work with arbitrary numbers of dimensions. [ghstack-poisoned]

ghstack-source-id: 805300c Pull Request resolved: #44550

ghstack-source-id: 805300c Pull Request resolved: pytorch#44550

Part of the `torch.fft` work (gh-42175). This adds n-dimensional transforms: `fftn`, `ifftn`, `rfftn` and `irfftn`. This is aiming for correctness first, with the implementation on top of the existing `_fft_with_size` restrictions. I plan to follow up later with a more efficient rewrite that makes `_fft_with_size` work with arbitrary numbers of dimensions. [ghstack-poisoned]

ghstack-source-id: 4fee26e Pull Request resolved: pytorch#44550

Part of the `torch.fft` work (gh-42175). This adds n-dimensional transforms: `fftn`, `ifftn`, `rfftn` and `irfftn`. This is aiming for correctness first, with the implementation on top of the existing `_fft_with_size` restrictions. I plan to follow up later with a more efficient rewrite that makes `_fft_with_size` work with arbitrary numbers of dimensions. [ghstack-poisoned]

codecov · 2020-09-20T03:05:54Z

Codecov Report

❗ No coverage uploaded for pull request base (gh/peterbell10/10/base@c70e6fe). Click here to learn what that means.
The diff coverage is n/a.

@@                    Coverage Diff                    @@
##             gh/peterbell10/10/base   #44550   +/-   ##
=========================================================
  Coverage                          ?   68.08%           
=========================================================
  Files                             ?      393           
  Lines                             ?    50765           
  Branches                          ?        0           
=========================================================
  Hits                              ?    34564           
  Misses                            ?    16201           
  Partials                          ?        0

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c70e6fe...9bccee0. Read the comment docs.

mruberry · 2020-09-22T04:14:30Z

Hey @peterbell10, this is looking great. Let me know when I should review it again and if there's anything I can help with. As mentioned on the next PR: 1.7's branch cut is 8 days away, so I'd like to get this in within the next week.

peterbell10 · 2020-09-22T12:25:44Z

@mruberry this is ready for another look.

Part of the `torch.fft` work (gh-42175). This adds n-dimensional transforms: `fftn`, `ifftn`, `rfftn` and `irfftn`. This is aiming for correctness first, with the implementation on top of the existing `_fft_with_size` restrictions. I plan to follow up later with a more efficient rewrite that makes `_fft_with_size` work with arbitrary numbers of dimensions. [ghstack-poisoned]

mruberry

Awesome work as usual, @peterbell10!

Part of the `torch.fft` work (gh-42175). This adds n-dimensional transforms: `fftn`, `ifftn`, `rfftn` and `irfftn`. This is aiming for correctness first, with the implementation on top of the existing `_fft_with_size` restrictions. I plan to follow up later with a more efficient rewrite that makes `_fft_with_size` work with arbitrary numbers of dimensions. [ghstack-poisoned]

peterbell10 · 2020-09-22T17:19:40Z

Fixed a few small doc errors which weren't updated properly after copy-pasting.

mruberry · 2020-09-23T12:33:39Z

test/test_spectral_ops.py

+
+    @skipCPUIfNoMkl
+    @skipCUDAIfRocm
+    @unittest.skipIf(not TEST_NUMPY, 'NumPy not found')


You may need @onlyOnCPUAndCUDA here and on another test below. Would you rebase to test the XLA build?

Noticed some of the fft tests were missing this as well. Added and rebased on viable/strict.

@mruberry

Thanks. XLA build looks good.

Part of the `torch.fft` work (gh-42175). This adds n-dimensional transforms: `fftn`, `ifftn`, `rfftn` and `irfftn`. This is aiming for correctness first, with the implementation on top of the existing `_fft_with_size` restrictions. I plan to follow up later with a more efficient rewrite that makes `_fft_with_size` work with arbitrary numbers of dimensions. Differential Revision: [D23846032](https://our.internmc.facebook.com/intern/diff/D23846032) [ghstack-poisoned]

facebook-github-bot · 2020-09-24T06:19:28Z

@mruberry merged this pull request in 6a2e9eb.

Summary: Between September 25 and September 27, approximately half an hour was added to the running time of `pytorch_linux_xenial_cuda10_2_cudnn7_py3_gcc7_test`. Judging from the CircleCI data, it looks like the majority of the new time was added by the following PRs: - #44550 - #45298 I'm not sure what to do about #44550, but it looks like #45298 increased the `N` for `TestForeach` from just 20 to include both 30 and 300. This PR would remove the 300, decreasing the test time by a couple orders of magnitude (at least when running it on my devserver), from over ten minutes to just a few seconds. Pull Request resolved: #45785 Reviewed By: malfet Differential Revision: D24094782 Pulled By: samestep fbshipit-source-id: 2476cee9d513b2b07bc384de751e08d0e5d8b5e7

torch.fft: Multi-dimensional transforms

8db6ed9

[ghstack-poisoned]

peterbell10 requested review from ebetica, goldsborough and yf225 as code owners September 11, 2020 13:02

This was referenced Sep 11, 2020

Cleanup tracer handling of optional arguments #43009

Closed

Add optional string support to native_functions schema #43010

Closed

Add one dimensional FFTs to torch.fft namespace #43011

Closed

peterbell10 marked this pull request as draft September 11, 2020 13:02

pytorchbot added the open source label Sep 11, 2020

Update on "torch.fft: Multi-dimensional transforms"

c41210b

[ghstack-poisoned]

Update on "torch.fft: Multi-dimensional transforms"

0b3d6a9

[ghstack-poisoned]

peterbell10 added a commit that referenced this pull request Sep 11, 2020

torch.fft: Multi-dimensional transforms

c9a9356

ghstack-source-id: bc92ef4 Pull Request resolved: #44550

peterbell10 added a commit that referenced this pull request Sep 12, 2020

torch.fft: Multi-dimensional transforms

3132139

ghstack-source-id: 16ea948 Pull Request resolved: #44550

peterbell10 added a commit that referenced this pull request Sep 12, 2020

torch.fft: Multi-dimensional transforms

14df591

ghstack-source-id: 805300c Pull Request resolved: #44550

peterbell10 added a commit to peterbell10/pytorch that referenced this pull request Sep 17, 2020

torch.fft: Multi-dimensional transforms

f8eb89b

ghstack-source-id: 805300c Pull Request resolved: pytorch#44550

This was referenced Sep 17, 2020

Deprecate old fft functions #44876

Closed

torch.fft helper functions #44877

Closed

peterbell10 added a commit to peterbell10/pytorch that referenced this pull request Sep 19, 2020

torch.fft: Multi-dimensional transforms

e1aacd3

ghstack-source-id: 4fee26e Pull Request resolved: pytorch#44550

peterbell10 added 2 commits September 19, 2020 15:28

mruberry added this to the 1.7.0 milestone Sep 22, 2020

mruberry approved these changes Sep 22, 2020

View reviewed changes

This was referenced Sep 22, 2020

codegen: Improve array default handing #45163

Closed

torch.fft: Two dimensional FFT functions #45164

Closed

mruberry reviewed Sep 23, 2020

View reviewed changes

peterbell10 added 3 commits September 23, 2020 14:52

facebook-github-bot closed this in 6a2e9eb Sep 24, 2020

facebook-github-bot added the merged label Sep 24, 2020

facebook-github-bot deleted the gh/peterbell10/10/head branch September 27, 2020 14:15

samestep mentioned this pull request Oct 2, 2020

Use smaller N to speed up TestForeach #45785

Closed

mruberry added the Merged label Oct 28, 2020

gchhablani mentioned this pull request Sep 29, 2021

Update FNet Fourier Transform huggingface/transformers#13752

Draft

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

torch.fft: Multi-dimensional transforms #44550

torch.fft: Multi-dimensional transforms #44550

Uh oh!

peterbell10 commented Sep 11, 2020 •

edited

Loading

Uh oh!

dr-ci bot commented Sep 11, 2020 •

edited

Loading

Uh oh!

codecov bot commented Sep 20, 2020 •

edited

Loading

Uh oh!

mruberry commented Sep 22, 2020

Uh oh!

peterbell10 commented Sep 22, 2020

Uh oh!

mruberry left a comment

Uh oh!

peterbell10 commented Sep 22, 2020

Uh oh!

mruberry Sep 23, 2020

Uh oh!

peterbell10 Sep 23, 2020

Uh oh!

mruberry Sep 23, 2020

Uh oh!

facebook-github-bot commented Sep 24, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

torch.fft: Multi-dimensional transforms #44550

torch.fft: Multi-dimensional transforms #44550

Uh oh!

Conversation

peterbell10 commented Sep 11, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dr-ci bot commented Sep 11, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

💊 CI failures summary and remediations

Uh oh!

codecov bot commented Sep 20, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

mruberry commented Sep 22, 2020

Uh oh!

peterbell10 commented Sep 22, 2020

Uh oh!

mruberry left a comment

Choose a reason for hiding this comment

Uh oh!

peterbell10 commented Sep 22, 2020

Uh oh!

mruberry Sep 23, 2020

Choose a reason for hiding this comment

Uh oh!

peterbell10 Sep 23, 2020

Choose a reason for hiding this comment

Uh oh!

mruberry Sep 23, 2020

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Sep 24, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

peterbell10 commented Sep 11, 2020 •

edited

Loading

dr-ci bot commented Sep 11, 2020 •

edited

Loading

codecov bot commented Sep 20, 2020 •

edited

Loading